Glossary
Browser use
Agent tools that read and act on a web page, without taking over a full desktop.
Browser use is a set of tools that let a model read a web page and act on it. The work stays in the browser. A full desktop, with every local app, is computer use.
Anthropic's computer-use page says the browser use tool reads and acts on the page itself and does not need a full desktop environment. OpenAI's computer-use guide separates BrowserToolset, for page text, controls, and tabs, from ComputerToolset, for screenshots and mouse or keyboard actions. Both still sit inside a harness that runs the calls and applies the permissions.
The longer guide is Can an AI agent use your computer?.